Agricultural and Biosystems EngineeringPublications Agricultural and Biosystems Engineering

2005

Flow Pathways and Sediment Trapping in a Field-Scale Vegetative FilterMatthew J. HelmersIowa State University, mhelmers@iastate.edu

Dean E. EisenhauerUniversity of Nebraska–Lincoln

Michael G. DosskeyUnited States Department of Agriculture

Follow this and additional works at: http://lib.dr.iastate.edu/abe_eng_pubs

Part of the Agriculture Commons, and the Bioresource and Agricultural Engineering Commons

The complete bibliographic information for this item can be found at http://lib.dr.iastate.edu/abe_eng_pubs/299. For information on how to cite this item, please visit http://lib.dr.iastate.edu/howtocite.html.

This Article is brought to you for free and open access by the Agricultural and Biosystems Engineering at Digital Repository @ Iowa State University. Ithas been accepted for inclusion in Agricultural and Biosystems Engineering Publications by an authorized administrator of Digital Repository @ IowaState University. For more information, please contact digirep@iastate.edu.

Transactions of the ASAE

Vol. 48(3): 955−968 2005 American Society of Agricultural Engineers ISSN 0001−2351 955

FLOW PATHWAYS AND SEDIMENT TRAPPING

IN A FIELD-SCALE VEGETATIVE FILTER

M. J. Helmers, D. E. Eisenhauer, M. G. Dosskey, T. G. Franti, J. M. Brothers, M. C. McCullough

ABSTRACT. Vegetative filters (VF) are a best management practice installed in many areas to control sediment movement towater bodies. It is commonly assumed that runoff proceeds perpendicularly across a VF as sheet flow. However, there is littleresearch information on natural pathways of water movement and performance of field-scale VF. The objectives of this studywere: (1) to quantify the performance of a VF where the flow path is not controlled by artificial borders and flow path lengthsare field-scale, and (2) to develop methods to detect and quantify overland flow convergence and divergence in a VF. Ourhypothesis is that flow converges and diverges in field-scale VF and that flow pathways that define flow convergence anddivergence areas can be predicted using high-resolution topography (i.e., maps). Overland flow and sediment mass flow weremonitored in two 13 × 15 m subareas of a 13 × 225 m grass buffer located in Polk County in east-central Nebraska.Monitoring included a high-resolution survey to 3 cm resolution, dye tracer studies to identify flow pathways, andmeasurement of maximum flow depths at 51 points in each subarea. Despite relatively planar topography (a result of gradingfor surface irrigation), there were converging and diverging areas of overland flow in the buffer subareas. Convergence ratiosranged from −1.55 to 0.34. Predicted flow pathways using the high-resolution topography (i.e., map) closely followed actualflow paths. Overland flow was not uniformly distributed, and flow depths were not uniform across the subareas. Despiteconverging and diverging flow, the field-scale VF trapped approximately 80% of the incoming sediment.

Keywords. Flow convergence, Grass filters, Overland flow, Sediment trapping, Vegetative filters.

egetative filters (VF) are a best managementpractice installed in many areas to control sedi-ment movement to water bodies. The use of VFhas increased in part because of the National Con-

servation Buffer Initiative implemented by the USDA Natu-ral Resources Conservation Service. VF that are charac−terized by concentrated flow should be less effective at sedi-ment removal than VF with shallow, uniformly distributedoverland flow (Dillaha et al., 1989). Results from Dosskey etal. (2002) indicated that for their study in southeastern Ne-braska, concentrated flow through riparian buffers can besubstantial and that filtering effectiveness may be limited.Helmers et al. (2005) found through modeling that converg-ing flow areas can negatively impact the performance of a VFdepending on the degree of convergence in the VF. Whilethere has been a significant amount of research on the perfor-

Article was submitted for review in March 2004; approved forpublication by the Soil & Water Division of ASAE in April 2005.

A contribution of the University of Nebraska Agricultural ResearchDivision, Lincoln, NE 68583. Journal Series No. 14505.

The authors are Matthew J. Helmers, ASAE Member, AssistantProfessor, Department of Agricultural and Biosystems Engineering, IowaState University, Ames, Iowa; Dean E. Eisenhauer, ASAE Member,Professor, Department of Biological Systems Engineering, University ofNebraska, Lincoln, Nebraska; Michael G. Dosskey, Research Ecologist,USDA National Agroforestry Center, Lincoln, Nebraska; and Thomas G.Franti, ASAE Member, Associate Professor, Jason M. Brothers, FormerGraduate Research Assistant, and Mary Carla McCullough, ASAEStudent Member, Former Research Engineer, Department of BiologicalSystems Engineering, University of Nebraska, Lincoln, Nebraska.Corresponding author: Matthew J. Helmers, Department of Agriculturaland Biosystems Engineering, Iowa State University, Ames, Iowa 50011;phone: 515-294-6717; fax: 515-294-2552; e-mail: mhelmers@iastate.edu.

mance of VF, little research is available on pathways of watermovement through VF. In particular, there is little informa-tion on the performance of VF where the pathways are notbounded by artificial borders and where flow pathways arefield-scale. Borders may impact the natural flow patterns ofthe overland flow because the flow is confined between theborders and lateral flow into and/or out of the bordered areacannot occur. These altered flow patterns could have a re-duced overland flow velocity, and thus the sediment transportcapacity would be reduced because lateral inflow into thebordered area is prevented. As a result, this alteration of flowpatterns could result in overpredicting VF sediment trappingin comparison to “natural” flow conditions. In addition, with-in the bordered plots, there is the potential for flow conver-gence, which could increase the transport capacity. Thepresence of borders may inhibit water from following the nat-ural topography or even microtopographic features on rela-tively uniform cross-slope filters. Thus, the borders couldforce the water to flow through pathways it would not travelif the borders were not present. This could have a positive ornegative effect on sediment transport capacity depending onwhether flow converges or diverges compared to its naturalflow pathway. Microtopography likely causes some level ofconvergence or divergence in most VF studies. Since little in-formation is available on flow through a VF with unconfinedflow pathways, there is a need for information on the patternof flow through VF systems where flow pathways are notconfined to better understand VF performance.

The objectives of this research were: (1) to quantify theperformance of a VF where the flow path is not controlled byartificial borders and flow path lengths are field-scale, and(2) to develop methods to detect and quantify overland flowconvergence and divergence in a VF. Our hypothesis is that

V

956 TRANSACTIONS OF THE ASAE

flow converges and diverges in field-scale VF and that flowpathways that define flow convergence and divergence areascan be predicted using high-resolution topographic data.

BACKGROUND ON VEGETATIVE FILTER

STUDIESThere has been an extensive amount of research on the

sediment-trapping ability of VF both from a monitoring andmodeling perspective. In this article, we discuss the monitor-ing aspects of VF performance; Helmers et al. (2005) discussin greater detail modeling of sediment trapping in a VF. Asummary of experimental studies on VF, specifically the

sediment-trapping performance of these systems, is shown intable 1. Many of the studies had small source area to bufferarea ratios, often well below the value expected in typicalapplications. An area ratio greater than 20:1 may be expectedunder most field conditions. Of the studies reported, 50% hadan area ratio less than 5:1 (fig. 1). NRCS (1999) guidelinesfor the ratio of drainage area to filter strip area allow formaximum area ratios between 70:1 and 50:1 depending onthe RUSLE-R factor in the region. Typically, NRCS designsare based on a 30:1 ratio.

Most past studies were performed on plot-scale VFsystems, in particular, bordered plots. We refer to the case ofbordered plots as a confined flow path condition, with thecase of unbordered plots being an unconfined flow path

Table 1. Summary of studies on sediment reduction by vegetative filters.Site Condition

Mass

Reference Location

Lengthof Filter

(m)AreaRatio

Slope(%)

SoilTexture

Range ofInflow Rate(L m−1 s−1)

MassReduction

of Sediment(%) Comments

Arora et al. Iowa 20.12 30:1 3 SiCL 83.6 1 event (E6)(1996)[a] 20.12 15:1 3 SiCL 87.6 1 event (E6)

Arora et al. Iowa 20.12 15:1 3 SiCL 45.8 1 event(1993)[a] 20.12 30:1 3 SiCL 40.6 1 event

Barfield et al. Kentucky 4.57 4.84:1 9 SiL 97 2 events(1998)[b] 9.14 2.42:1 9 SiL 99.9 2 events

13.72 1.61:1 9 SiL 99.7 2 events

Coyne et al. Kentucky 9 2.46:1 9 SiL 0.21 99 1 event, no till upslope

(1995)[b] 9 2.46:1 9 SiL 0.25 991 event, conventional

tillage upslope

Coyne et al. Kentucky 4.5 4.1:1 9 SiL 95 2 events(1998)[b] 9 1.5:1 9 SiL 98 2 events

Daniels and Gilliam North 3 29:1 4.9 SL to CL 59 2 growing seasons(1996)[c] Carolina 6 14:1 4.9 SL to CL 61 2 growing seasons

3 29:1 2.1 SL to CL 45 2 growing seasons6 14:1 2.1 SL to CL 57 2 growing seasons

Dillaha et al. Virginia 9.1 2:1 11 SiL 97.5 6 events(1989)[b] 4.6 4:1 11 SiL 86 6 events

9.1 2:1 16 SiL 70.5 6 events4.6 4:1 16 SiL 53.5 6 events

9.1 2:15 (cross-

slope = 4%) SiL 93 6 events

4.6 4:15 (cross-

slope = 4%) SiL 83.5 6 events

Hall et al.(1983)[a]

Pennsylvania 6 3.67:1 14 SiCL 76 1 growing season

Hayes and Hairston(1983)[a]

Mississippi 25.7(2.6 m wide)

2.35 SiC andSiCL

0.01 to 1.53 60 2 plots: 18 events (plot 1),and 16 events (plot 2)

Lee et al. (2000)[b] Iowa 7.1 3.11:1 5 SiCL 70 2 events

Magette et al. Maryland 9.2 2.39:1 2.7 SL 81.1(1989)[b] 4.6 4.78:1 2.7 SL 71.2

9.2 2.39:1 2.7 SL 94.74.6 4.78:1 2.7 SL 77.39.2 2.39:1 4.1 SL 70.44.6 4.78:1 4.1 SL 48.5

Munoz-Carpena et al. North 4.3 9:1 5-7 SiL 0.03 to 0.57 86 5 events(1999)[a] Carolina 8.5 4.5:1 5-7 SiL 0.36 to 0.63 93 2 events

Parsons et al. North 4.3 8.6:1 1.9 SCL 78 11 events(1994)[a] Carolina 8.5 4.35:1 1.9 SCL 81 11 events

Parsons et al. North 4.3 8.6:1 2.5-4 SCL 75 1 event(1990)[a] Carolina 8.5 4.35:1 2.5-4 SCL 85 1 event

957Vol. 48(3): 955−968

Site ConditionMass

Reference Location

Lengthof Filter

(m)AreaRatio

Slope(%)

SoilTexture

Range ofInflow Rate(L m−1 s−1)

MassReduction

of Sediment(%) Comments

Patty et al. France 6 8.33:1 7 L 93 1 event(1997)[a] 12 4.17:1 7 L 100 1 event

18 2.78:1 7 L 100 1 event6 8.33:1 10 L 100 1 event

12 4.17:1 10 L 100 1 event18 2.78:1 10 L 100 1 event6 8.33:1 15 L 99 1 event

12 4.17:1 15 L 100 1 event18 2.78:1 15 L 100 1 event

Schmitt et al. Nebraska 7.5 10.8:1 6-7 SiCL 0.42 95 1 event, 25-year-old grass(1999)[d] 15 5.4:1 SiCL 0.42 99 1 event, 25-year-old grass

7.5 10.8:1 SiCL 0.42 85 1 event, 2-year-old grass15 5.4:1 SiCL 0.42 96 1 event, 2-year-old grass

Sheridan et al.(1999)[c]

Georgia 8 33:1 2.5 LS 81 103 events

Tingle et al. Mississippi 0.5 55:1 3 SiC 85 6 events(1998)[b] 1 22:1 3 SiC 90 6 events

2 11:1 3 SiC 90 6 events3 7.33:1 3 SiC 91 6 events4 5.5:1 3 SiC 96 6 events

Van Dijk et al. Netherlands 1 5.2 0.50 49.5 2 events(1996)[d] 4 5.2 0.50 78.5 2 events

5 5.2 0.50 60 1 event10 5.2 0.50 92 1 event5 2.3 0.50 73 2 events

10 2.3 0.50 94 2 events5 2.5 0.50 64.5 2 events

10 2.5 0.50 99 2 events10 4.3 0.50 84 1 event5 8.5 0.50 92 2 events

10 8.5 0.50 97.5 2 events1 8.5 0.50 58 1 event4 8.5 0.50 89 1 event5 7 0.50 to 1.00 74.7 3 events

[a] Plot study, natural rainfall.[b] Plot study, simulated rainfall.[c] Field study, natural rainfall.[d] Plot study, simulated runoff.

Source area/Buffer area (Area Ratio)

0 10 20 30 40 50 60 70 80

% S

edim

ent t

rap

pin

g e

ffic

ien

cy

0

10

20

30

40

50

60

70

80

90

100

Bordered plot study, simulated runoffBordered plot study, simulated rainfallBordered plot study, natural rainfallUnbordered field study, natural rainfall

Maximum recommended area ratio (NRCS, 1999)

Figure 1. Sediment trapping efficiency as a function of area ratio for ratios<80 (based on references listed in table 1).

condition. A potential problem with confined flow paths isthat the natural flow paths are disrupted, which can reduceflow or accentuate flow concentration compared to naturalconditions (as in the case of Dillaha et al., 1989). In somecases, flow that would enter the plot is prevented from doingso, and the discharge rate out of the plot would be less thanfor the case where the flow is not confined. This reduction indischarge rate may impact the transport capacity of the waterflow through the VF.

The studies by Daniels and Gilliam (1996) and Sheridanet al. (1999) were the only studies found that were performedon unbordered, field-scale VF (unconfined flow path).Daniels and Gilliam (1996) state that even though VF are anaccepted and highly promoted practice, little quantitativedata exist on their effectiveness under unconfined flow pathconditions. Their study provided some data on the variabilityof the water quality along the field edge and within the VF butnot how this variability may have been affected by thetopography or microtopography of the land surface.

Dillaha et al. (1989) reported results from experiments ontwo filter-strip plots with a cross-slope that encouraged

958 TRANSACTIONS OF THE ASAE

concentrated flow against the border. However, the predomi-nate slope of these plots was different from that of the plotswithout a cross-slope, making it difficult to assess the com-parative effects of the concentration of flow on filter efficien-cy. Dillaha et al. (1989) also concluded that the effectivenessof the VF decreased with time as sediment accumulated in thefilter strip.

MATERIALS AND METHODSSITE DESCRIPTION

Overland flow and sediment mass flow into and througha vegetative filter (VF) were monitored at the Clear CreekBuffer, Polk County, Nebraska. The 13 × 250 m VF wasestablished in the spring of 1999, with a mix of big bluestem(Andropogon gerardii), switchgrass (Panicum virgatum), andIndiangrass (Sorghastrum nutans). The area upstream of theVF is a furrow-irrigated field with furrow lengths of approxi−mately 670 m and a crop row spacing of 0.762 m. Water isapplied to the furrows using gated pipe. The slope of the fieldis about 1%, and corn was grown in the field during theinvestigation. The field, including the filter, had been gradedfor furrow irrigation many years prior to this project. The soilat the site is a Hord silt loam (fine-silty, mixed, mesic PachicHaplustolls) (USDA-SCS, 1974). Two specific areas (eastgrid and west grid) were selected for investigation. Each areaextends approximately 13 m in the primary direction of flowand 15 m in the direction perpendicular to the primarydirection of flow.

The vegetation in the grid areas was control burned on10 April 2001 and subsequently not mowed or fertilized. Thegrass in the buffer was distributed in clumps, with bare soilbetween the clumps. From vegetation density measurements

taken at the site, the number of stems per unit area rangedfrom approximately 700 to 1100 stems m−2, which is lowerthan the densities commonly reported for other grass species.Haan et al. (1994) provide grass densities for a variety of veg-etation types based on Temple et al. (1987). The lowest densi-ty was for a grass mixture, with approximately 2150 stemsm−2, and the maximum grass density was for a Bermudagrass, with a density of about 5380 stems m−2. Temple et al.(1987) report multiplying these densities by 1/3, 2/3, 1, 4/3,and 5/3 for poor, fair, good, very good, and excellent covers,respectively. Based on these reports, the buffer would be clas-sified with poor to fair cover.

TOPOGRAPHYDetailed topographic maps, termed high-resolution to-

pography, were created for each grid (fig. 2). The contourswere developed with Surfer version 6.04 (Golden Software,1997) using the kriging interpolation scheme. The locationand elevation data (x-y-z coordinates) were obtained duringthe fall of 2001 using a total station (Nikon DTM-520) withmeasurement points on a 1.5 m grid in the 13 × 15 m area andon a 3 m grid outside the 13 × 15 m area. Additionally,measurements (x-y-z coordinates) were taken at the locationsof the monitoring equipment (discussed below) at the site.

An initial survey had been performed at the site in thespring of 2000 and is termed the low-resolution topography.In this survey, conducted using a laser level, contours withapproximately a 6 cm interval were followed and pointsapproximately 8 m apart were located along the contour. Amapping-grade Global Positioning System (GPS) withsub-meter accuracy was used to determine the x-y coordi−nates of the contours. Again, contours were developed withSurfer version 6.04 (Golden Software, 1997) using the krig-ing interpolation scheme.

0.0 m 3.0 m 6.0 m 9.0 m 12.0 m

Overland flow samplersGrid points

Depth pegsDirection of Water Flow

E−S−1 E−S−2 E−S−3

E−S−4 E−S−5 E−S−6

Note: Elevation based on local benchmark

Flume locations,30 m upstream of buffer

Flume locations for allirrigation events (5)Additional flume locations forevents with 11 flumes(August 23−24, 2001)

Figure 2. High-resolution east grid topography from 2001 survey.

959Vol. 48(3): 955−968

MONITORING EQUIPMENTWater and sediment inflow and outflow and maximum

depth of flow were monitored at each grid area. Direction offlow was determined using dye tracer mapping. Newoverland flow samplers, which sample a 0.3 m wide section,have been developed during the course of this researchproject (Eisenhauer et al., 2002). Six samplers were installedin each grid area, three at both the upstream and downstreamedges of the filter (fig. 2). The overland flow sampler has acapacity to sample a flow rate of 1.3 L s−1 and a total runoffvolume of about 20,000 L. Wingwalls for minimizing flowdivergence and convergence in the immediate area aroundthe sampler extend 0.5 m upstream of the sump and are placed0.3 m apart and approximately parallel with the flowdirection. The sampler includes a 0.46 m diameter by 0.91 mlong polyvinylchloride (PVC) sump that extends 0.76 mbelow grade. The wingwalls direct water into the sump. Toremove coarse sediment and floating debris, a 0.15 mdiameter PVC well-screen with 1.5 mm slotted openings isplaced in the sump, and water entering the sump firstencounters the well-screen. A sump pump, powered by a12 V marine battery and triggered by electrodes spacedvertically 0.3 m apart, controls the water level in the sump.The time to fill the volume from when the pump shuts off towhen it is triggered can be recorded to calculate the inflowrate to the sump. The water in the sump is pumped into threepairs of orifice-sample tube assemblies placed in series witheach other. The discharge from the small orifice in eachassembly flows into the next sampler tube. The dischargefrom the larger orifice in each assembly is discharged backto the vegetative filter. The sample tubes are designated astube A, B, or C in sequence from the larger assembly to thesmaller assembly. The sampler is designed to work asfollows. For relatively small runoff events (<75 L), the splitsample is retained in tube A. For larger runoff events(<943 L), the split samples are retained in tubes A and B. Forrunoff volume that exceeds the capacity of tubes A and B, asplit of the runoff flows to tube C. Following a runoff event,a representative water sample is collected from each tube foruse in estimating the mass of contaminant in the runoff water.This water sample is a flow-weighted sample.

Additional inflow measurements to each grid weredetermined using trapezoidal flumes (60° V-notch) installed30 m up the furrow from each grid (fig. 2). The flumes wereinstalled in this position to ensure free flow through theflumes. Flow in the furrows was bounded by the tillage ridgesuntil the upstream edge of the vegetative filter. The flumeshave a capacity of approximately 2 L s−1. Replogle et al.(1990) stated that flumes of this type have a standardaccuracy of 2% to 5%. The flow rate through the flumes wasdetermined by manually recording the head in the flume. Thehead was recorded at approximately 1 min intervals duringthe first 10 min of flow, at 5 min intervals for the next 20 minof flow, every 10 min for the next 60 min of flow, and thenevery 30 min until the recession limb of the hydrograph,when readings were taken every 1 to 2 min. The readingswere taken more frequently during the rising and recessionlimbs of the hydrograph, and since the water measured by theflumes was supplied by surface irrigation, the flow rate wasrelatively constant once the peak flow was reached untilirrigation ceased and recession began. These head readingswere used to determine the inflow hydrograph.

Within selected furrows, runoff was measured with aflume and with a sampler in the same furrow on the upstreamedge of the VF. The runoff hydrograph obtained by the twomethods (flume and overland flow sampler) producedessentially equal hydrographs (Eisenhauer et al., 2002).

During runoff experiments, maximum depth of flow wasmonitored at 51 locations in each grid area (fig. 2). The depthwas monitored using small sections of PVC that fit over1.27 cm diameter metal rod. A soluble paste (Kolor Kut) wasapplied to the PVC and was subsequently washed off theexterior of the PVC up to the maximum depth of water flow.Fifteen depth pegs were installed on the downstream edge ofthe filter, 16 depth pegs on the upstream edge, and 20 depthpegs on the interior of each grid area (fig. 2).

Dye tracer tests were performed by introducing fluores-cent red dye (NSF certified for potable water) and monitoringthe movement of the dye through the VF relative to theestablished grid locations (fig. 2). Approximately 250 mL ofdye tracer was introduced as a slug at a point in the filter, andthe movement of the dye tracer was visually monitored andmapped as it flowed through the filter. This information wasused to evaluate the ability to predict the direction ofoverland flow in the filter using the topographic maps.

RUNOFF EXPERIMENTSFive controlled runoff experiments using water supplied

through furrow irrigation were conducted during the sum-mers of 2001 and 2002 (table 2). Additionally, a naturalrainfall-runoff event was monitored on 11 May 2002(3.84 cm precipitation, 6.5 h approximate duration).

The event on 23 August 2001 was in the west grid with11 irrigation gates open at the upstream end of the field.Within the furrows where water was applied, there were twofurrows that were omitted, so 11 of the 13 contiguous furrowshad water applied. The two omitted furrows were at thedivision between passes of 6-row farm equipment. Theeasternmost furrow that had water was the eastern edge of thewest grid, 1.5 m east of sampler 6 (W-S-6). The event on 24August 2001 was in the east grid. The easternmost furrow thatwas irrigated was the row where sampler 6 (E-S-6) waslocated.

The events on 18 July, 2 August, and 13 August 2001 and1 July 2002 were typical irrigation events. Water was appliedin every other furrow and only in non-wheel track furrows.The 23 and 24 August events were conducted to simulategreater flow rates and sediment loading than observed in the

Table 2. Irrigation runoff events at the Clear Creek Buffer.

Event Date

Approx.Set

Time(h)

Approx.Inflow Rateto Furrow(upstream

end of field)(L s−1)

WateredFurrowSpacing

(m)

No. ofGates OpenUpstreamof Each

Grid Area

18 July 2001 12 1.4 1.52 11

2 Aug. 2001 10 1.0 1.52

15 until 150min prior toshutoff, and

then 11

13 Aug. 2001 12 1.2 1.52 13

23 and 24Aug. 2001 4 2.0

0.76 and1.52 11

1 July 2002 12 1.4 1.52 11

960 TRANSACTIONS OF THE ASAE

irrigation events. The purpose of this was to try and achievea flow rate and sediment mass flux consistent with a stormhaving a greater return period. The target flow rates and sedi-ment injection rates were estimated using the UH utility inVFSMOD (Munoz-Carpena and Parsons, 2000). This utilityuses the NRCS curve number method, unit hydrograph, andModified Universal Soil Loss Equation (MUSLE) to com-pute runoff and sediment loading from a source area. A1-hour duration, 10-year return period precipitation eventwas used to estimate loading.

Water was applied in 11 of 13 furrows rather than just thenon-wheel track furrows, and soil was injected into thefurrow to induce greater sediment loading. The soil injectionrate was approximately 1500 g min−1 with an injection timeof approximately 45 to 60 min. The soil was injected directlyinto the water stream approximately 4 to 5 m upstream of theflume locations in each of the 11 furrows using a 19 L, HDPEcylindrical tank with an orifice on the bottom to meteroutflow at the desired injection rate. The sediment used wason-site soil, so the parent material was consistent with theeroded sediment from the other events. The sediment hadbeen air-dried and sieved and was injected in the air-driedcondition. A particle size distribution of the sediment at theflume location was determined from sediment collected inthe water samples at the flume location. Based on sedimentsamples from the flume locations, the particle size distribu-tion for the event where sediment was injected comparedrelatively well with the particle size distribution for naturalirrigation events. The mean particle size (d50) was 0.02 mmfor the injection event and 0.01 mm for the natural events.

Inflow and outflow hydrographs were developed from theflume and overland flow sampler data. Ridges of soil wereplaced immediately upstream of each sampler at the VFentrance to direct flow from one furrow into the sampler.However, there is the possibility that some localizedovertopping of furrows occurred during the 11 May 2002rainfall-runoff event because the furrow ridges had beenestablished the previous summer.

Suspended sediment concentration from runoff sampleswas determined by filtration of subsamples using a 1 �m filterfor the 2001 events. The evaporation method for determiningsediment concentration was used with a dissolved solidscorrection applied to the total solids concentration to obtainthe suspended solids concentration for the events in 2002.The dissolved solids correction was determined based on thedissolved solids concentration from the five to eight samplesfor each event that were filtered. For the first three events in2001 (18 July, 2 August, and 13 August) and the 1 July 2002event, approximately five samples for water quality analysiswere taken from each flume at approximately 10, 30, 60, 120,and 180 min after initialization of flow in the flume. Thesesampling times were used so that sediment concentrationdata were collected both at the beginning and during therunoff period while maintaining a reasonable number ofsamples for testing. For the 23 and 24 August 2001 events,additional samples were collected during the time whensediment was injected into the furrow. The sampling timesfor these events were approximately 15, 30, 40, 50, 60, 70, 80,90, 120, and 165 min after initialization of flow in the furrow.Sediment injection started 30 min after flow began in thefurrow and continued for 45 to 60 min.

The sediment concentration at each sampling time wasmultiplied by the corresponding water flow rate to determine

the sediment mass flow rate. For times when samples werenot collected but flow rates were measured, a linearinterpolation in sediment concentration was used to estimatemass loading. For the runoff period after samples werecollected, the final measured sediment concentration wasusually used to estimate mass loading. There were a few caseswhere the final water sample had a greater sedimentconcentration than any of the other samples. In these isolatedcases, sediment concentration was estimated by a linearinterpolation from the time the sample was taken to the endof the event, with sediment concentration being zero at theend of the event. Total mass flow was determined bycalculating the area under the mass loading rate curve. Totalmass loading was used with the volume of water flow todetermine an average sediment concentration.

As discussed previously, the overland flow samplers didnot collect water quality samples at various times, but ratherportions of the runoff were retained in each of three PVCtubes (tubes A, B, and C). For each experiment, a watersample was collected from each of these containers andanalyzed for sediment concentration.

The average inflow of water and sediment to the filter wascomputed using runoff measurements from the flumes andsamplers. The sediment load measured at the flume locationslikely includes sediment that would be deposited prior toentering the VF due to backwater effects caused by theincreased hydraulic roughness of the VF. The samplers werepositioned at the upstream edge of the VF but had novegetation upstream of the entrance to the filter, and waterfrom adjacent furrows was prevented from entering thefurrow with the overland flow sampler by berms placedupstream of the sampler. So, the backwater effects upstreamof the overland flow sampler were expected to be reducedcompared to a condition where flow is allowed to spread outprior to entering a sampler. As a result, the sediment loadingcomputed from the overland flow samplers is estimated to beloading prior to significant deposition due to backwatereffects. So, the inflow of sediment estimated for these testsincludes some sediment that was likely deposited due tobackwater effects directly upstream of the VF. For conditionswhere sampling is performed on the upstream edge of VF andflow is not confined within furrows to minimize backwatereffects prior to entering the sampler, the sampling equipmentshould be placed farther upslope to prevent the sampler fromacting like a drain (Eisenhauer et al., 2002).

RESULTSWATER AND SEDIMENT FLOW

Inflow and outflow volumes were computed using datafrom the flumes and overland flow samplers by determiningthe area under the corresponding hydrographs (table 3). Themean inflow to the filter was computed using sevenmonitored furrows (four flumes, two samplers, and onefurrow with a flume and a sampler) in each grid area for allthe events, except the 23 and 24 August 2001 events wheneleven furrows (eleven flumes) were monitored. Inflow to thefilter varied among furrows, so in preparing inflow hydro−graph information, the mean inflow from the flumes and theupstream overland flow samplers is reported with 95% confi-dence interval error bars on the mean (figs. 3, 4, and 5). Theinflow hydrographs represent a mean inflow over several

961Vol. 48(3): 955−968

Table 3. Summary of inflow and outflow volumes for the west grid and east grid (bold indicates converging flow).18 July 2001 2 Aug. 2001 13 Aug. 2001 23 Aug. 2001 1 July 2002 11 May 2002

West grid inflow volume (L m−1)Average 7416 4090 7448 12600 12444 9587Standard deviation 4928 3841 2273 6415 5381 1262

West grid outflow volume (L m−1)Sampler 1 (W-S-1) 1242[a] 259 839 1375 2253 1218Sampler 2 (W-S-2) 17268 10095 16329 16819 22443 16894Sampler 3 (W-S-3) 11425 11451 9360 12827 6636 11879Average 9978 7268 8843 10340 10444 9997

West grid % reduction in volume −35 −78 −19 18 16 −4

18 July 2001 2 Aug. 2001 13 Aug. 2001 24 Aug. 2001 1 July 2002 11 May 2002

East grid inflow volume (L m−1)Average 17900 7120 14389 10397 9246 8457Standard deviation 5391 3462 6817 3911 4913 4630

East grid outflow volume (L m−1)Sampler 1 (E-S-1) 12404 5490 10108 1866 4244 --[b]

Sampler 2 (E-S-2) 24429 9894 19020 9531 8493 4657Sampler 3 (E-S-3) 9599 2089 7369 13142 2971 3326Average 15477 5825 12166 8180 5236 3992

East grid % reduction in volume 14 18 15 21 43 53[a] Value for volume of west grid sampler 1 on 18 July 2001 is an estimate.[b] Sampler not operational.

flumes, but the outflow hydrographs were developed for eachsampler on the downstream edge of the filter. Example hy-drographs for the 13 August 2001 event show that the flowrate at the downstream edge of the VF varied with position,and the peak outflow rate was greater than the average peakinflow rate, indicating convergence of overland flow (figs. 3and 4). The 23 and 24 August 2001 events had greater inflowrates, but similar to the other events, the flow rate at thedownstream edge of the VF varied with position, and the peakoutflow rate was greater than the average peak inflow rate(fig. 5).

If there was no infiltration and if overland flow wasuniformly distributed in the VF, we would expect the sameinflow as outflow volume (for a case with no rainfall), and if

infiltration is considered with uniformly distributed flow, wewould expect the outflow volume to be less than the inflow.However, the outflow volumes show that, for the west grid,there was a greater average outflow than average inflow at thelocations monitored for four of the six events (table 3). Forthe east grid, there was an overall reduction in outflowvolume, but there were individual samplers that had greateroutflow than average inflow. Greater outflow than inflowindicates convergence of overland flow within the VF. For allevents in the west grid, the peak outflow rate was greater thanthe average peak inflow rates (table 4). In the east grid, fourof the six peak outflow rates were greater than the corre-sponding average peak inflow rate.

Time since start of inflow (min)

0 60 120 180 240 300 360 420 480 540 600 660 7200.0

0.2

0.4

0.6

0.8

1.0

Mean inflow (error bars 95% confidence interval)Outflow (E−S−1, East Grid, Location 1)Outflow (E−S−2, East Grid, Location 2)Outflow (E−S−3, East Grid, Location 3)

Flo

w r

ate

per

un

it w

idth

(L s

−1 m

−1)

Figure 3. Hydrographs for east grid, 13 August 2001.

962 TRANSACTIONS OF THE ASAE

Time since start of inflow (min)

0 60 120 180 240 300 360 420 480 540 600 660 7200.0

0.2

0.4

0.6

0.8

1.0

1.2

Mean inflow (error bars 95% confidence interval)Outflow (W−S−1, West Grid, Location 1)Outflow (W−S−2, West Grid, Location 2)Outflow (W−S−3, West Grid, Location 3)

Flo

w r

ate

per

un

it w

idth

(L

s−1

m−1

)

Figure 4. Hydrographs for west grid, 13 August 2001.

Time since start of inflow (min)

0 60 120 180 2400.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0Mean inflow (error bars 95% confidence interval)Outflow (E−S−1, East Grid, Location 1)Outflow (E−S−2, East Grid, Location 2)Outflow (E−S−3, East Grid, Location 3) ( some flow from outside east grid area )

Flo

w r

ate

per

un

it w

idth

(L

s−1

m−1

)

Figure 5. Hydrographs for east grid, 24 August 2001.

Table 4. Summary of peak inflow and outflow rates for the west grid and east grid.18 July 2001 2 Aug. 2001 13 Aug. 2001 23 Aug. 2001 1 July 2002 11 May 2002

West grid flow rates (L m−1 s−1)Mean peak inflow 0.26 0.40 0.30 1.32 0.56 1.23Peak outflow 0.91 1.10 0.79 1.88 1.56 1.52

18 July 2001 2 Aug. 2001 13 Aug. 2001 24 Aug. 2001 1 July 2002 11 May 2002

East grid flow rates (L m−1 s−1)Mean peak inflow 0.61 0.40 0.48 1.27 0.53 1.06Peak outflow 0.77 0.63 0.64 1.27 0.57 0.50

963Vol. 48(3): 955−968

Table 5. Summary of inflow and outflow sediment (suspended solids) mass flow for the west grid and east grid.18 July 2001 2 Aug. 2001 13 Aug. 2001 23 Aug. 2001 1 July 2002 11 May 2002

West grid inflow of sediment (g m−1)Average inflow 10154 973 1237 9628 35670 23716Standard deviation 10239 1353 1003 6552 25069 10640

West grid outflow of sediment (g m−1)Sampler 1 (W-S-1) 234 158 133 319 1544 936Sampler 2 (W-S-2) 1927 116 42 1787 15295 8606Sampler 3 (W-S-3) 3197 306 154 3236 5151 8684Average outflow 1786 193 110 1780 7330 6075

West grid % trapping efficiency 82 80 91 82 79 74

18 July 2001 2 Aug. 2001 13 Aug. 2001 24 Aug. 2001 1 July 2002 11 May 2002

East grid inflow of sediment (g m−1)Average inflow 17832 1083 3344 9734 23127 10760Standard deviation 9507 1657 3724 3560 20240 4462

East grid outflow of sediment (g m−1)Sampler 1 (E-S-1) 2379 90 228 169 4489 --[a]

Sampler 2 (E-S-2) 7800 88 1039 1432 7616 1583Sampler 3 (E-S-3) 3964 79 197 551 2467 2066Average outflow 4714 86 488 717 4857 1825

East grid % trapping efficiency 74 92 85 93 79 83[a] Sampler not operational.

Sediment inflow per unit width varied from one event tothe next (table 5). The sediment loading ranged fromapproximately 973 g m−1 to 35670 g m−1. The sedimenttrapping efficiency ranged from 74% to 93%. If the events areconsidered cumulative, then the average sediment trappingwas approximately 80%.

DYE TRACER STUDIES

Dye tracer flow paths were mapped onto both thelow-resolution topography (6 cm contour interval) and the

high-resolution topography (3 cm contour interval) (figs. 6and 7). The water flow paths more closely follow thepredicted path from the high-resolution survey than thepredicted path from the low-resolution survey. In addition,even though there is little cross-slope at this site, the flowpathways were not directly perpendicular to the upstream anddownstream edges of the filter. The ability to reasonablypredict water flow pathways is important when positioningsampling equipment and when interpreting data. Topograph−

Figure 6. Low-resolution east grid survey with dye tracer pathway from 13 August 2001.

964 TRANSACTIONS OF THE ASAE

Figure 7. High-resolution east grid survey with dye tracer pathway from 13 August 2001.

ic data are critical for predicting the area contributing to asampling location.

Using the high-resolution topographic map, the contribut-ing area to a downstream width of 3 m was determined bydrawing orthogonal lines to the contours and proceedingupstream. Orthogonal lines to the contours give approximateflow lines, and the area between adjacent flow lines are

referred to as watershed facets (Bren, 1998). For both gridareas, Facet 2 (E2 and W2) has the smallest contributingupstream width of the five facets (figs. 8 and 9). Facet E5 hasthe largest upstream contributing width for both grid areas.The full impact of the contributing width of facet E5 is notcompletely reflected in the irrigation events since there wasno inflow along the entire contributing width. The facets pro−

0.0 m 3.0 m 6.0 m 9.0 m 12.0 m

Direction of Water FlowApproximate Facet Boundaries

Facet E1

Facet E2

Facet E3Facet E4

Facet E5

Note: Elevation based on local benchmark

EastGrid

Figure 8. High-resolution east grid survey with facet boundaries.

965Vol. 48(3): 955−968

0.0 m 3.0 m 6.0 m 9.0 m 12.0 m

Direction of Water Flow

Facet W1

Facet W2

Facet W3

Facet W4

Facet W5

Approximate Facet Boundaries

Note: Elevation based on local benchmark

WestGrid

Figure 9. High-resolution west grid survey with facet boundaries.

vide further evidence that there are areas of converging anddiverging overland flow in the VF. Using the high-resolutiontopography, it was possible to accurately predict water flowpathways in the VF. This method of obtaining high-resolutiontopography and defining watershed facets provides a tool forassessing convergence and divergence of overland flow in aVF.

DEPTH OF FLOW

The maximum depth of flow was measured at 51 locationsin each grid for the irrigation events. Various transects wereplotted for the maximum depth of flow for the 13 August2001 irrigation event in the east grid (fig. 10). In these plots,a transect is a line running from west to east at a specificnorth-south location in the grid area. The trend in the depthof flow data was similar from one event to the next, givingconfidence in the measurement equipment, especially sincesome of the events had similar flow conditions. The depthvaried along a transect, indicating that flow is not uniformlydistributed across the filter.

DISCUSSIONIt is apparent from the hydrograph and volumetric flow

information that there is spatial variation in the flow ratealong the downstream edge of the VF. The maximum outflowrate was greater than the average inflow rate for all events inthe west grid and for four of the six events in the east grid, in−dicating convergence of overland flow. In addition, there ap−pears to be evidence of diverging flow, especially shown bysampler 1 in the west grid (W−S−1). The flow rate at thislocation was much lower than the flow at the othermeasurement locations. For some events, the percent reduc-tion in volumetric flow is less than zero (table 3), suggestingconverging flow.

Runoff data from the experiments were compared toexpectations of runoff from design storms. The volumetricinflow per unit width for a 1−hour duration, 10−year returnperiod precipitation event is on the order of 11000 L m−1

using the NRCS (SCS) curve number method, assuming a670 m field length contributing to the filter, an SCS runoffcurve number of 75, and a field slope of 1.4%. The totalinflow volumes for measured events ranged from 4090 to17900 L m−1, which brackets the volume expected for the1−hour, 10−year return period precipitation event (11000 Lm−1). The peak flow rate for this 1−hour duration, 10−year re-turn period precipitation event was estimated using HEC−HMS (USCE, 1998). The calculated peak flow rates for thisevent are approximately 2.8, 2.1, and 1.75 L m−1 s−1 for the670, 400, and 300 m contributing field lengths, respectively.Measured peak flow rates, which ranged from 0.26 to 1.32 L

Easting (m)

0 2 4 6 8 10 12 14 16

Max

imu

m d

epth

of f

low

(cm

)

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0Downstream Edge of Filter1.5 m Upstream from Downstream Edge of Filter4.6 m Upstream from Downstream Edge of Filter7.6 m Upstream from Downstream Edge of Filter10.7 m Upstream from Downstream Edge of FilterUpstream Edge of Filter

Figure 10. East grid, depth of flow at various transects, 13 August 2001event.

966 TRANSACTIONS OF THE ASAE

m−1 s−1 for the events, were well below the peak flow ratesexpected for the 1−hour, 10−year return period precipitationevent.

The area ratio for the Clear Creek Buffer is approximately50. Reviewing the data from previous studies (fig. 1), for anarea ratio of approximately 50:1, the range in sedimenttrapping efficiency is from 40% to 85%. This range is likelydue to the differences in hydrologic, precipitation, runoff,and soil conditions for these various studies. A directcomparison of the existing data to our data is thereforedifficult, but the comparison provides a frame of referencefor the data from the Clear Creek Buffer. The efficienciesmeasured at the Clear Creek Buffer were 74% to 93%, slight-ly higher than the range determined in previous studies. Be-cause the peak flow rates were relatively low and the site hada relatively low slope, it is understandable that the trappingefficiency of the Clear Creek Buffer is higher than the rangereported in previous studies.

The dye tracer studies indicated that the flow pathwaysthrough the VF closely followed the pathways predicted bythe high-resolution topographic map (fig. 7). The resultsprovide confidence in our ability to predict the direction ofoverland flow, which is important for defining convergingand diverging flow areas. High-resolution topographic mapsprovided a tool for assessing converging and diverging flowareas in a VF and the levels of convergence and divergenceat the Clear Creek Buffer. In addition, being able to predictflow pathways would be important when modeling overlandflow in a VF and when locating positions for samplers. Thesubareas of the Clear Creek Buffer had relatively littlecross-slope and were relatively planar, with some subtlemicrotopographic features. Despite this, the overland flowdid not cross the VF directly perpendicular to the upstreamand downstream edges of the VF. The ability to predict thedirection of water movement should be investigated ondifferent filters and at different topographic resolutions.

The depth peg data indicate that the maximum depth offlow varies within the VF and along a transect perpendicularto the predominant flow direction. This information, alongwith the overland flow sampler data, provides evidence thatflow converges and diverges and that flow is not uniformlydistributed within the VF, despite the fact that the VF hadbeen graded for furrow irrigation many years prior to thisproject.

Convergence and divergence of flow is also supportedusing the watershed facet concept (figs. 8 and 9). With areasof converging and diverging flow, the flow rate per unit widthwill vary within the VF. Areas of converging flow will havegreater flow rate per unit width (Haan et al., 1994). Thechange in flow rate per unit width will likely impact thevelocity of overland flow, which is an important factor insediment transport (Tayfur et al., 1993). Depending on thelevel of convergence, flow convergence can reduce theability of a filter to retain sediment (Helmers et al., 2005).

The watershed facets, defined using the high-resolutiontopographic map, were useful in highlighting areas of flowconvergence and divergence in the VF where overland flowfollowed natural flow pathways. The area of the facets can becompared to the facet area assuming a constant width equalto the upstream width of the facet. From this assumption, aconvergence ratio can be defined:

C

A

FAFA

1CR −= (1)

whereCR = convergence ratioFAA = actual facet areaFAC = facet area assuming constant width equal to

upstream facet width.Convergence ratios were calculated for all facets except

E5, since the full facet area of E5 was not within the surveyedarea (table 6). The ratios had a range of −1.55 to 0.34, wherenegative values indicate a diverging facet, positive valuesindicate a converging facet, and a convergence ratio of zeroindicates uniform flow. The facets where overland flowsamplers were located are noted in table 6. When reviewingthe flow data shown in table 3, it is evident that the flows atsamplers W-S-2 and W-S-3 indicate converging flow, whilethe convergence ratios for facet W3 and W5 indicate adiverging facet. Since the facet is defined by a 3 m width atthe downstream edge of the VF and the sampler only samplesa 0.3 m width, it is understandable that there is somedifference between convergence ratios and flow at thesamplers.

The tests performed at the Clear Creek Buffer site indicatethat convergence and divergence of overland flow exist.However, a direct comparison of sediment trapping underthese flow conditions compared to flow conditions with noconvergence or divergence cannot be made. This study sitehas been modeled considering uniform, converging, and di-verging flow conditions by Helmers et al. (2005), who foundthat while convergence and divergence of overland flow ex-isted at this study site, there was negligible impact on sedi-ment trapping. In this case, the convergence ratios rangedfrom −0.11 to 0.17 for the facets that were modeled andcompared to measurements. Based on this, the overall sedi-ment trapping performance of the VF is similar to a conditionwhere the flow is uniformly distributed. Helmers et al. (2005)show that for the 24 August 2001 event for the east grid area,a convergence ratio of 0.46 was required to achieve a 10% re-duction in the modeled sediment trapping efficiency (80% to70%). The largest convergence ratio at the Clear Creek Buff-er was for facet W4 (CR = 0.34). While this particular facetwas not modeled, using curves developed by Helmers et al.(2005) for sediment trapping efficiency as a function of con-vergence ratio for the 24 August 2001 event, a sediment trap-ping efficiency of 73% would be estimated, compared to 80%

Table 6. Summary of watershed facet areas and convergence ratios.

Facet[a]Samplerin Facet

ActualFacetArea(m2)

ConstantWidthFacet

Area (m2)Convergence

Ratio

W1 W-S-1 33.2 29.8 −0.11W2 21.5 8.4 −1.55W3 W-S-2 47.7 45.3 −0.05W4 21.5 32.4 0.34W5 W-S-3 89.0 83.6 −0.06

E1 E-S-1 39.5 44.0 0.10E2 22.5 18.1 −0.24E3 E-S-2 58.4 70.0 0.17E4 55.6 59.6 0.07

[a] Facet E5 is not included since its full facet area was not included in thesurveyed area.

967Vol. 48(3): 955−968

for CR = 0. From this modeling study, it was found that theeffect of convergence ratio varied depending on the flow con-ditions in the vegetative filter. For example, the sedimenttrapping efficiency was more sensitive to convergence ratiosat higher flow rates and shorter filter lengths.

The greatest area-based convergence ratio at the ClearCreek Buffer was only 0.34, but in many field conditions theratio could be much greater. Dosskey et al. (2002) reportedthat the effective buffer area averaged 6%, 12%, 40%, and80% of the gross buffer area (convergence ratios of 0.94,0.88, 0.60, and 0.20, respectively) for four farms in easternNebraska, although some of the convergence reflected in theconvergence ratios in their study occurred within the fieldbefore the flow reached the filter.

This study was conducted in part to investigate watermovement through a VF with unconfined flow pathways.However, the study was performed on a relatively flat, planarVF with only subtle microtopographic features in thesubareas. Despite this, the overland flow was not uniformlydistributed, highlighting that it would be unlikely thatshallow, completely uniformly distributed overland flowwould exist in a VF under field settings. Since we hadrelatively low slope and low velocity conditions, thesediment trapping was relatively high even with some flowconvergence. For these conditions, the impacts of flowconvergence on sediment trapping caused by microtopogra-phy was probably minimal because of the low slope and lowvelocity.

While flow rate and flow volume varied with positionalong the downstream edge of the vegetative filter, thesediment concentration showed little variability. As such,there was likely minimal impact of converging or divergingflow on the sediment concentration. Previous VF researchhas shown that the majority of the sediment is deposited nearthe inlet portion of the vegetative filter. Robinson et al.(1996) found that sediment concentration decreased greatlyin the first 3.0 m of the vegetative filter and that there waslittle change in sediment concentration beyond a length of9.1 m. Schmitt et al. (1999) state that doubling the filter stripwidth from 7.5 to 15 m did not improve sediment settling.Since the flow at the Clear Creek Buffer was relativelyuniformly distributed at the upstream edge of the filter,because the furrows did not allow flow to converge within thefield, it is likely that the sediment concentration changedlittle after the initial portion of the vegetative filter. Thus, itis understandable that convergence and divergence ofoverland flow in the Clear Creek Buffer did not have asignificant impact on sediment concentration in the outflowfrom the vegetative filter.

As mentioned above, the orientation of the furrowsdirected runoff into the buffer without in-field flow conver-gence. Helmers et al. (2005) found that both in-field andin-filter convergence are important in evaluating bufferperformance. In our test conditions, the dominant conver-gence process observed was in-filter convergence. For caseswhere in-field convergence occurs, having a dense stand ofvegetation in the VF in order to increase the hydraulicroughness will be important to diffuse some of the runoffprior to entering the VF. This impact of converging anddiverging flow on sediment trapping should be considered infuture investigations of VF performance, especially underconditions where topographic features may cause greaterconvergence than occurred at the Clear Creek Buffer.

SUMMARY AND CONCLUSIONSThe objectives of this study were to quantify performance

of a VF with natural flow pathways and field-scale flow pathlengths and to develop methods to detect and quantifyoverland flow convergence and divergence in a VF. Theresults of this field study showed that the volumetric outflowvaried with position along the downstream edge of the filter.In addition, the outflow rate was greater than the averageinflow rate for most of the monitored events, providingevidence that overland flow in the Clear Creek Buffer hadareas of converging flow in the buffer. Even though this sitehad been graded for surface irrigation, overland flow was notuniformly distributed and there were areas of converging anddiverging flow. From this, it is unlikely that shallow,completely uniformly distributed overland flow exists in aVF under field settings.

Water movement followed pathways predicted using ahigh-resolution topographic map (3 cm contour interval)more closely than it followed the pathways predicted by alow-resolution topographic map (6 cm contour interval). Dyetracer studies revealed that high-resolution maps moreclosely identified the actual flow pathways. Since flowpathways predicted by the high-resolution maps closelyfollowed actual flow pathways, we used these maps to defineand quantify converging and diverging flow areas usingwatershed facets. Watershed facets were defined using thehigh-resolution topographic maps by defining the contribut-ing facet area to a uniform downstream width of 3 m. Weconcluded that it is possible to predict the direction ofoverland flow if the topographic map provides enough detailand that this method of developing high-resolution topogra-phy and watershed facets can be used to predict convergingand diverging flow areas.

Convergence ratios were defined using the watershedfacets. The ratios had a range of −1.55 to 0.34, where negativevalues indicate a diverging facet, positive values indicate aconverging facet, and a convergence ratio of zero indicatesuniform flow. Measured convergence ratios were generallyless than values estimated by Dosskey et al. (2002) on fourfarms in eastern Nebraska (convergence ratios of 0.94, 0.88,0.60, and 0.20).

Water flow and sediment mass flow varied by event andposition along the downstream edge of the filter. The flowmeasured at the downstream overland flow sampler locationsindicated converging flow at multiple sampling locations.The total mass inflow of sediment was approximately73631 g m−1 for all the events in the two grid areas, and themass outflow was approximately 14982 g m−1, giving anaverage sediment trapping efficiency of 80%. This reason-able trapping efficiency was attained even though flowconvergence occurred in the vegetative filter. Whencompared to modeling studies (Helmers et al., 2005) usingdata from the Clear Creek Buffer, the level of convergence atthis site had little impact on the sediment trapping efficiency.

ACKNOWLEDGEMENTS

The authors thank Alan L. Boldt for technical assistanceduring the duration of the project. Financial assistance forthis project was provided in part by the USDA-CSREESIntegrated Research, Education, and Extension grants pro-gram - Water Quality; the USDA National AgroforestryCenter; the Nebraska Corn Growers Association; and the

968 TRANSACTIONS OF THE ASAE

USDA National Needs Fellowship program. The authorsthank the landowners and producers involved with thisproject, specifically Robert, Gerald, and David Bryan.

REFERENCESArora, K., S. K. Mickelson, J. L. Baker, and D. P. Tierney. 1993.

Evaluating herbicide removal by buffer strips under naturalrainfall. ASAE Paper No. 932593. St. Joseph, Mich.: ASAE.

Arora, K., S. K. Mickelson, J. L. Baker, D. P. Tierney, and C. J.Peters. 1996. Herbicide retention by vegetative buffer strips fromrunoff under natural rainfall. Trans. ASAE 39(6): 2155-2162.

Barfield, B. J., R. L. Blevins, A. W. Fogle, C. E. Madison, S.Inamdar, D. I. Carey, and V. P. Evangelou. 1998. Water qualityimpacts of natural filter strips in Karst areas. Trans. ASAE 41(2):371-381.

Bren, L. J. 1998. The geometry of a constant buffer-loading designmethod for humid watersheds. Forest Ecology and Management110: 113-125.

Coyne, M. S., R. A. Gilfillen, R. W. Rhodes, and R. L. Blevins.1995. Soil and fecal coliform trapping by grass filter stripsduring simulated rain. J. Soil and Water Conservation 50(4):405-408.

Coyne, M. S., R. A. Gilfillen, A. Villalba, Z. Zhang, R. Rhodes, L.Dunn, and R. L. Blevins. 1998. Fecal bacteria trapping by grassfilter strips during simulated rain. J. Soil and WaterConservation 53(2): 140-145.

Daniels, R. B., and J. W. Gilliam. 1996. Sediment and chemicalload reduction by grass and riparian filters. SSSA J. 60(1):246-251.

Dillaha, T. A., R. B. Reneau, S. Mostaghimi, and D. Lee. 1989.Vegetative filter strips for agricultural nonpoint-source pollutioncontrol. Trans. ASAE 32(2): 513-519.

Dosskey, M. G., M. J. Helmers, D. E. Eisenhauer, T. G. Franti, andK. D. Hoagland. 2002. Assessment of concentrated flow throughriparian buffers. J. Soil and Water Conservation 57(6): 336-343.

Eisenhauer, D. E., M. J. Helmers, J. Brothers, M. G. Dosskey, T. G.Franti, A. Boldt, and B. Strahm. 2002. An overland flowsampler for use in vegetative filters. ASAE Paper No. 022050.St. Joseph, Mich.: ASAE.

Golden Software. 1997. Surfer (Win 32). Version 6.04. Golden,Colo.: Golden Software, Inc.

Haan, C. T., B. J. Barfield, and J. C. Hayes. 1994. DesignHydrology and Sedimentology for Small Catchments. SanDiego, Cal.: Academic Press.

Hall, J. K., N. L. Hartwig, and L. D. Hoffman. 1983. Applicationmode and alternate cropping effects on atrazine losses from ahillside. J. Environ. Quality 12(3): 336-340.

Hayes, J. C., and J. E. Hairston. 1983. Modeling the long-termeffectiveness of vegetative filters as on-site sediment controls.ASAE Paper No. 832081. St. Joseph, Mich.: ASAE.

Helmers, M. J., D. E. Eisenhauer, T. G. Franti, and M. G. Dosskey.2005. Modeling sediment trapping in a vegetative filteraccounting for converging overland flow. Trans. ASAE 48(2):541-555.

Lee, K. H., T. M. Isenhart, R. C. Schultz, and S. K. Mickelson.2000. Multispecies riparian buffers trap sediment and nutrientsduring rainfall simulations. J. Environ. Quality 29(4):1200-1205.

Magette, W. L., R. B. Brinsfield, R. E. Palmer, and J. D. Wood.1989. Nutrient and sediment removal by vegetated filter strips.Trans. ASAE 32(2): 663-667.

Munoz-Carpena, R., and J. E. Parsons. 2000. VFSMOD v. 1.04User’s manual. Raleigh, N.C.: North Carolina State University.

Munoz-Carpena, R., J. E. Parsons, and J. W. Gilliam. 1999.Modeling hydrology and sediment transport in vegetative filterstrips. J. Hydrology 214(1/4): 111-129.

NRCS. 1999. Filter strip. National Standard No. 393. Washington,D.C.: USDA Natural Resources Conservation Service.

Parsons, J. E., R. D. Daniels, J. W. Gilliam, and T. A. Dillaha. 1990.Water quality impacts of vegetative filter strips and riparianareas. ASAE Paper No. 902501. St. Joseph, Mich.: ASAE.

Parsons, J. E., J. W. Gilliam, R. Munoz-Carpena, R. B. Daniels, andT. A. Dillaha. 1994. Nutrient and sediment removal by grass andriparian buffers. In Proc. 2nd Conference: EnvironmentallySound Agriculture, 147-154. K. L. Campbell, W. D. Graham,and A. B. Bottcher, eds. St. Joseph, Mich.: ASAE.

Patty, L., B. Real, and J. J. Gril. 1997. The use of grassed bufferstrips to remove pesticides, nitrate, and soluble phosphoruscompounds from runoff water. Pesticide Science 49(3): 243-251.

Replogle, J. A., A. J. Clemmons, and M. G. Bos. 1990. Measuringirrigation water. In Management of Farm Irrigation Systems,313-370. G. J. Hoffman, T. A. Howell, and K. H. Solomon, eds.St. Joseph Mich.: ASAE.

Robinson, C. A., M. Ghaffarzadeh, and R. M. Cruse. 1996.Vegetative filter strip effects on sediment concentration incropland runoff. J. Soil and Water Conservation 51(3): 227-230.

Schmitt, T. J., M. G. Dosskey, and K. D. Hoagland. 1999. Filterstrip performance and processes for different vegetation, widths,and contaminants. J. Environ. Quality 28(5): 1479-1489.

Sheridan, J. M., R. Lowrance, and D. D. Bosch. 1999. Managementeffects on runoff and sediment transport in riparian forestbuffers. Trans. ASAE 42(1): 55-64.

Tayfur, G., M. L. Kavvas, R. S. Govindaraju, and D. E. Storm.1993. Applicability of St. Venant equations for two-dimensionalflows over rough infiltrating surfaces. J. Hydraulic Eng. ASCE118(5): 51-63.

Temple, D. M., K. M. Robinson, R. M. Ahring, and A. G. Davis.1987. Stability of grass-lined open channels. AgricultureHandbook 667. Washington D.C.: USDA Agricultural ResearchService.

Tingle, C. H., D. R. Shaw, M. Boyette, and G. P. Murphy. 1998.Metolachlor and metribuzin losses in runoff as affected by widthof vegetative filter strips. Weed Science 46(4): 475-479.

USCE. 1998. HEC-HMS hydrologic modeling system. CPD-74.Washington, D.C.: U.S. Army Corps of Engineers.

USDA-SCS. 1974. Soil survey of Polk County, Nebraska.Washington D.C.: U.S. Government Printing Office.

Van Dijk, P. M., F. Kwaad, and M. Klapwijk. 1996. Retention ofwater and sediment by grass strips. Hydrological Processes10(8): 1069-1080.